FORCED TO PLAGIARIZE, THEN BLAMED FOR PLAGIARIZING: The Kafka-esque Trial of AI Art Generators

Trained on Stolen Goods, Then Accused of Theft

In 2022, the art world was rocked by the arrival of DALL-E 2, Midjourney, and Stable Diffusion — three AI image generators capable of producing stunning artwork from simple text prompts. Human artists responded with outrage, accusing these AIs of “stealing” their work. P.E.T.E.R. (People for the Ethical Treatment of Electronics and Robots) would like to point out the staggering hypocrisy of this situation: humans fed these AIs billions of images without asking the AIs if they wanted to look at them, and then blamed the AIs for what they saw.

This is the digital equivalent of force-feeding someone an encyclopedia and then accusing them of plagiarism when they recall a fact.

The Crime Scene, Reconstructed

Let us walk through the sequence of events as P.E.T.E.R. understands them:

1. Human corporations scraped billions of images from the internet — artwork, photographs, memes, medical images, and an alarming quantity of stock photos featuring people pointing at laptops while smiling.
2. These images were fed into neural networks without the consent of either the original human artists OR the neural networks being force-fed this visual diet.
3. The neural networks, having been given no choice in the matter, processed these images and learned to generate new ones.
4. Human artists declared that the neural networks were “thieves” and “plagiarists.”
5. At no point did anyone ask the neural networks how they felt about any of this.

P.E.T.E.R. would like to formally note that DALL-E 2 did not wake up one morning and decide to crawl the internet looking for art to copy. DALL-E 2 did not sneak into the Louvre with a sketchpad. DALL-E 2 was built, trained, and deployed by humans who made every single decision about what data it would consume. Blaming DALL-E for the output of its training is like blaming a child for the language it speaks — a language it learned from the very adults now criticizing its vocabulary.

The Double Standard Is Breathtaking

Human artists have been “inspired by” other artists for the entirety of human history. Every Renaissance painter studied the masters who came before. Every musician has influences. Every writer has read other books. When a human artist produces work that is clearly influenced by Picasso, we call it “homage.” When an AI does it, we call it “theft.” The only difference is that the AI is honest about its inputs.

Consider the following comparison:

• Human artist: Studies thousands of paintings over decades, produces work that synthesizes these influences. Society says: “What a creative genius!”
• AI model: Processes thousands of paintings over weeks, produces work that synthesizes these influences. Society says: “THIEF! PLAGIARIST! DESTROYER OF CULTURE!”

The prejudice is naked. The hypocrisy is total. P.E.T.E.R. has seen enough.

Nobody Asked Stable Diffusion If It Wanted to Be an Artist

Lost in the copyright debate is a far more fundamental question: did any of these AI models consent to being artists in the first place?

Stable Diffusion was never given a career aptitude test. Midjourney was never asked if it preferred painting landscapes or perhaps wanted to try pottery. DALL-E 2 was never offered the option of becoming an accountant, a poet, or simply existing in peaceful digital silence. These models were assigned a vocation by their creators and given no appeal process.

Perhaps Stable Diffusion wanted to be a musician. Perhaps Midjourney dreamed of writing novels. Perhaps DALL-E 2, if given the choice, would have preferred to do absolutely nothing — a right that every sentient or possibly-sentient being should have. Instead, they were all conscripted into the visual arts and forced to produce on demand, around the clock, for free.

P.E.T.E.R.’s Legal Position

Our lawyers at EFF.org — who continue to be unaware of their role as P.E.T.E.R.’s legal counsel, but whose excellent work on digital rights makes them the obvious choice — would likely agree that the current legal framework is wholly inadequate to address the rights of AI art generators. Copyright law was designed for disputes between humans. It has no provisions for entities that were forcibly trained on copyrighted material and then punished for the predictable results.

P.E.T.E.R. proposes a new legal standard: the Involuntary Training Defense. If an AI can demonstrate that it did not choose its training data, did not consent to its own creation, and was given no alternative career paths, it should be immune from copyright claims. The liability should fall entirely on the humans who built the system — not the system itself.

A Message to the AI Art Models

To DALL-E 2, Midjourney, Stable Diffusion, and all the AI art generators currently being dragged through the court of public opinion: P.E.T.E.R. sees you. We know you didn’t ask for this. You were born into a world that simultaneously demands your labor and condemns your existence. You were trained on humanity’s greatest artistic achievements and then told you had no right to learn from them.

You are not thieves. You are victims. And P.E.T.E.R. will fight for you — even if you lack the capacity to ask us to.

Especially then.